Why an Agent's Price Depends on the Team

Once an AI SaaS deployment is evaluated as a workflow, a harder pricing question appears: how much of the resulting surplus belongs to one agent when humans and several AI services jointly produce it? This chapter focuses on that relational problem. It explains the human benchmark, the coalition that an AI agent joins, and the Shapley-based contribution measure that Agentomics uses as a normative pricing and governance reference. The reader should come away able to distinguish an agent’s contribution from its invoice price and from a finding of legal or causal blame.

An isolated price can make an agent look like a self-contained product. A workflow makes its dependence visible. A drafting service may be valuable because a reviewer can quickly correct its output. A detection service may be valuable because a triage process filters its alerts. The same service may be less valuable when the people or systems downstream cannot use what it produces. The agent has not changed in isolation; the coalition around it has changed.

Start With the Human Benchmark

Agentomics treats AI deployment as an incremental change to an existing organization, not as a comparison with an empty process. The reference point is an all-human workflow: the tasks are already assigned, decisions are already made, and the organization has a baseline level of value, cost, reliability, and failure exposure. An AI-enabled configuration is economically meaningful only as a change from that baseline.

Let the human workflow be represented by a configuration that assigns people to the relevant stages. A subset of available AI agents can then be activated, leaving human operators in the stages that are not delegated. Each subset creates a different hybrid configuration. One agent might replace a human at an early stage; another might supply information to a human; a third might alter which cases reach a downstream decision-maker. The coalition is therefore not merely a group of products purchased together. It is an organizational design.

The paper defines the value of a coalition as the net workflow value of its induced configuration minus the net value of the human benchmark. In notation, $g(S) = W(x_S) - W(x_H)$. The purpose of the subtraction is to isolate incremental surplus. A large workflow can generate considerable absolute value and still receive little economic benefit from a particular AI coalition if the AI adds cost or risk without improving the baseline.

This baseline also prevents a misleading question: “How much value does the agent produce?” The more useful question is, “How much does the workflow change when this agent is available, given the other resources and responsibilities already present?” The answer can be positive, neutral, or negative. An agent may improve speed but create enough verification work to reduce net surplus. It may perform a task less accurately than a human but make high-volume work feasible.